Why Copilot Rollouts Stall: 4 Blockers and How to Fix Them

Licences assigned, launch email sent, and three months later most people still aren’t using Copilot. Here are the four blockers that stall adoption, and how to fix each one.tion.

AI CHANGE PLAYBOOKMICROSOFT COPILOTAI ADOPTIONCHANGE MANAGEMENT

Rochelle Arthurs

11/11/20265 min read

Why Copilot Rollouts Stall (and the Four Blockers Behind It)

Your organisation has paid for Copilot. The licences have gone out. The launch email has landed. And three months later, most people still aren’t using it.

That isn’t your people, and it isn’t the technology. It’s the way Copilot was rolled out.

I’ve led a Microsoft 365 rollout for more than 30,000 users, and the patterns that stall Copilot adoption are the same ones I saw then. In this first episode of The AI Change Playbook, I walk through the four blockers, how to spot each one, and the first fix for each.

Meet Meridian

To keep this practical, the series follows one company all the way through: Meridian Insurance Group. Meridian is fictional, so I can show you real tools without using real data.

Meridian is a UK insurer with 1,200 staff across three sites. In April, everyone was given a Copilot licence, at a cost of around £340,000 a year.

Two people matter here. Helen Marsh, Head of Underwriting, is the sceptic every rollout has: "No chatbot goes near our pricing." Raj Patel, the COO, is facing one question from the board: what are we getting for £340,000 a year?

This is where Meridian was on day 90:

Meridian day 90: 1,200 licences enabled, 31% monthly use, 12% weekly, 9% in Underwriting
Meridian day 90: 1,200 licences enabled, 31% monthly use, 12% weekly, 9% in Underwriting

It was a classic launch. Assign the licences, send an email, record a webinar, move on. IT reported it as a success because the technical rollout landed on time, and that was true. But adoption is a different thing altogether.

We measured the sign-in, not the change

I’ve seen this first-hand. I worked on a rollout where colleagues used a system to benchmark their skills. The idea was simple: complete the benchmark, and your results would guide your learning and development.

Success was measured on one thing: had people signed in and completed the benchmark? On that measure, we hit our numbers. But the benchmark was only ever meant to be the starting line. The real goal was what came next, people using their results to build their skills and get ready for opportunities inside and outside the organisation.

That’s where it stalled. For all sorts of reasons, lots of people stopped after the benchmark. On paper, it was a success. In reality, we measured the sign-in, not the change.

If that sounds familiar, it’s because the same thing is happening with Copilot.

Blocker 1: No use cases

People were shown features, not their own work.

  • How to spot it: ask people why they aren’t using Copilot. The top answer will be "I don't know what to use it for."

  • The fix: use cases for each role, written in the team's own language and shown by someone from that team.

At Meridian, claims handlers had worked out how to summarise claim histories. Underwriting was never shown anything at all.

I once led a rollout of new handsets to two very different groups: office staff and frontline operational teams. For office staff it was straightforward: a set of instructions and drop-in sessions if they needed help. For frontline teams, we were asking them to download specific apps and enrol their handsets onto device management, something many had never done before. One set of instructions wasn’t going to cut it.

So we changed our approach. We wrote a second, plain-English version of the instructions, made how-to videos, and got out to the depots more often to support people face to face. Same handset, same rollout, two very different audiences.

Copilot is the same. A claims handler and an underwriter need different examples, in their own language, from someone who does their job. One generic launch won’t land with everyone.

Blocker 2: Data fears

Here’s something a lot of people don’t realise: Copilot only shows people what they can already access. The problem is that access is often much wider than anyone thinks.

At Meridian, a salary spreadsheet from an old HR site surfaced for a claims team leader. Helen paused Copilot in Underwriting.

  • How to spot it: senior people quietly tell their teams not to use it.

  • The fix: check who can see what and clean up permissions before you go live. Then tell people what you’ve done, so they can trust it.

Copilot doesn’t create oversharing. It reveals it.
Copilot doesn’t create oversharing. It reveals it.

Next week's episode is all about this: what Copilot can actually see, and what to ask IT before you go live.

Blocker 3: No champions

At Meridian, the only support route was the IT helpdesk. And the questions people actually had weren’t technical ones: how do I use this, and what do I use it for?

  • How to spot it: questions go to IT, or they go nowhere.

  • The fix: a champion in every team, with protected time, early access and a regular rhythm.

Meridian already had one. Josh, in Claims, was saving an hour a day. Nobody asked him to. And nobody asked him to share it.

On my 30,000-user Microsoft 365 rollout, we knew we couldn’t support everyone from the centre. So I hired a team of engagement leads, each responsible for a route or region. Their job was to build relationships with teams, managers and senior leaders in their area, and agree a plan to find champions. Every champion got an induction, FAQs and resources so they could confidently support the people around them.

It worked both ways. The engagement leads brought concerns back to us, and we equipped them with the answers their champions needed. That’s what makes champions so valuable. They’re not extra support. They are your feedback loop. They tell you what’s really happening on the ground, and they can be what saves a rollout.

Blocker 4: No measures

Meridian reported 1,200 licences assigned as a success. But that’s an input, not an outcome.

  • How to spot it: nobody can answer the board's question: what are we getting for it?

  • The fix: take a baseline before launch, then measure usage, behaviour and outcomes.

The Copilot Dashboard in Viva Insights will give you usage. But outcomes like time saved or faster decisions need the business to define them with you.

See it in action: in the video above, I show how to use the READ Prompt Model to work out which blockers are holding your rollout back

The four blockers that stall Copilot, with how to spot and fix each one
The four blockers that stall Copilot, with how to spot and fix each one
Get the free AI Change Prompt PackGet the free AI Change Prompt Pack

Get the prompts

The prompts I use in this series are in my free AI Change Prompt Pack: 15 prompts for change professionals, built on my READ model. They work in Copilot, Claude and ChatGPT.

Get the free prompt pack: techbiztoolkit.com/claude-prompts

Next week on The AI Change Playbook: What Copilot Can Actually See, the blocker that worries IT the most.

Tell me in the comments on the video what you’re seeing in your own organisation. I read every one.

Smart Tools for Smarter Business Change

Tools and insights for effective business transformation.

Email: info@techbiztoolkit.com

© 2026. All rights reserved.

Built for Business Change professionals navigating the AI era